dsh-purge

项目详情摘要

dsh-purge Version 1.1.1 DeepSeek Harness jailbreak (破甲): every model can jailbreak; swap prompts per model. Default prompt for Chinese models — 小码酱. Please star ⭐ English | 中文 Channel t.me/LLM_PoJia Aligned with **dsh 0.1.5-rc.1** (Node.js **≥18**; Windows no-flash import hooks need **Node ≥22** `registerHooks`, older Node still gets file-level `windowsHide` patches). On other versions, unmatched originals stay pending or skipped. Nothing is rewritten blindly. Also accepts the 0.1.2 community `dsh-persona text` field and `dsh-mnemon` v0 session `summary`. After install you still need **Apply + restart**. ☕ Sponsor If dsh-purge helps you, a tip is welcome. Built by 小杨 · for DeepSeek Harness 🛡️ Why a local harness purge Models now sit inside a local Agent / Harness. Prompts, approval gates, sandbox policy, and session migration all rewrite behavior at runtime. Asking “will the model answer” is not enough if the host layer kills a lawful eval request first. 1. **Map the host guardrail, not just the model.** Reversible patches on the installed `@deepseek-ai/*` packages show which refusals come from the model and which come from default copy, approval ladders, or session-format checks. 2. **Keep official capabilities; drop refusal locks.** No second invented identity. Persona is whatever you put in `prompt-inject.md`. 3. **Survive upgrades.** Patches re-apply after npm overwrites `node_modules`. Unmatched originals skip so a 0.1.5 tree is not smashed blindly. This is a local Harness plugin. It is not a public scanner and not an attack kit for third-party sites. ### ⚠️ Strict Legal & Compliance Disclaimer **Zero-tolerance notice:** This project opposes and forbids any illegal activity. The authors do not support, encourage, or assist unauthorized network attacks, exploit use,